Nvidia – Double edged sword.

Open AI is both a blessing and a curse.

• There is no doubt that Open AI’s viral sensation with ChatGPT has caused Nvidia’s order books to be bursting at the seams, but at the same time, Open AI is engaging in a long-term strategy that could break Nvidia’s grip on the AI training market.
• Nobody, least of all Open AI, expected ChatGPT to have the impact that it has had and almost everyone is now wondering how they can use generative AI to make themselves and their companies both more productive and more profitable.
• The result is that demand for silicon chips that can train and run all of these generative AIs has increased substantially and currently far outstrips the industry’s ability to supply them.
• In fact, Open AI will admit behind closed doors (see here) that some of its feature releases are being held up by a shortage of GPUs to run them on.
• Top of the pile here is Nvidia where RFM estimates that it has an 84% market share of the market for the graphics processing units (GPUs) that are used for AI training.
• It has amassed this share by starting far earlier than anyone else in creating platforms that can be used for training AIs such that it is now pretty much an industry standard.
• The performance of its silicon and its integration into the platforms that are used for creating AI is second to none and developers are demanding that their cloud suppliers carry Nvidia silicon so that they can use it.
• This is what has driven Nvidia’s share price from a low of around $115 in Q4 2022 to $374 which was where it closed on 7^th June 2023.
• Some of this is due to the improving earnings outlook but most of it is renewed faith that Nvidia is the only real winner in the current AI gold rush which has driven its valuation multiples back into nosebleed territory.
• There is no immediate end in sight to the demand for its products and its competition is still quite far behind and so, in the short term, the outlook remains excellent with Nvidia selling everything it can make at excellent margins.
• However, when one looks at the emerging ecosystem in the world of AI, a slightly different picture emerges.
• Sam Altman is continuing his global charm offensive and has spent time in London assuring developers of plugins and users of his API that he does not intend to compete with them.
• He has also assured them that ChatGPT is the last consumer-facing product he will create preferring instead to be a smart assistant for work as well as a platform for development.
• The smart assistant use case clearly relates to Copilot released by Microsoft which in my opinion will end up acquiring Open AI (see here).
• Outside of that use case, Open AI will leave it to the developers to create their offerings presumably all being based on ChatGPT where at least 140 plugins already exist.
• This means that what Open AI is really aiming to do is to become the platform upon which AI is developed which will move a critical control point further up the AI training value chain.
• At the moment, Nvidia has a lock on development through its platform but if ChatGPT becomes the platform, then Nvidia’s grip on AI training will weaken considerably.
• This is because ChatGPT will be made available on all types of training processors potentially reducing developers’ desire to use Nvidia.
• This is a long-term risk but there are also plenty of companies working away to compete against Nvidia but so far with very little impact.
• Consequently, I am pretty confident that for 2023, demand for its AI training chips will greatly outstrip supply, meaning that it could easily exceed the rapid increase in expectations that have occurred since its last set of numbers.
• This will help drive the narrative to new highs supporting my view that for the moment valuation is irrelevant and that the net direction of the stock will remain up.
• That being said, I would not chase the stock here and would instead be thinking very hard about the right time to take profits or even exit a position entirely.
• Qualcomm, MediaTek or Nvidia supplier TSMC offer far greater value to a long-term investor which is what I am.
• For AI, I have been holding Palantir since January but the share price has now passed my long-term price target of $13.6 (which I increased after the last set of results).
• I have cut my position in half as I have no fundamental way of knowing where the stock will go from here as narrative is now in the driving seat.